0016: Article 6 (Ramanujan's Pi formulas)


Ramanujan’s Pi Formulas with a Twist


By Tito Piezas III



Abstract:  A certain function related to Ramanujan’s pi formulas is explored at arguments k = {-½, 0, ½} and a conjecture will be given.


I. Introduction

II. Fundamental d with class number h(-d) = 1,2

III. Non-fundamental d with class number h(-d) = 1,2

IV. Higher class numbers h(-d) = 4,6

V.  Conclusion: Conjecture

(For an updated summary, also see Ramanujan's Pi Formulas and the Hypergeometric Function.)



I. Introduction


In his 1914 paper, “Modular Equations and Approximations to π”, Ramanujan gave 17 formulas for pi, one of which was,




with the Pochhammer symbol (a)n = (a)(a+1)(a+2)…(a+n-1).  He didn’t provide a proof, just stating that there were “corresponding theories”.  It was not until the 1980s work by the Chudnovsky and Borwein brothers that these formulas were placed on a rigorous footing.


In 2003, J. Guillera observed in his paper “Series closely related to Ramanujan formulas for pi” that, for k = ½,



It is easily seen that Ramanujan’s is just the case k = 0.  Guillera found that most of Ramanujan’s formulas (as well as those by the Chudnovskys, Borweins, and others) if modified in a similar manner, then evaluated to mln(x), where m and x are rationals. Aptly enough, there is still no proof that the relationships are true, though it holds for arbitrary thousands of decimal digits.  What the “corresponding theories” look like remains to be seen. 



II. Fundamental d with class number h(-d) = 1,2


In Guillera’s paper, the evaluation mln(x) had a fluctuating m.  To have a more well-behaved function, we will use the form,



for some constant k, discriminant d, Pochhammer products hp, and algebraic numbers {A,B,C}.  Define the hp as,



For k = 0, the formulas below neatly sum to,



However, for k = ½, then,


p = 1:



p = 2:



p = 3:



For p = 1, the denominators of the formulas are the exact values of the negated j-function j(τ), while for p = 2,3, these are for the related modular functions r2(τ), r3(τ).  (See article, Pi formulas and the Monster group).  Note the prime factorizations,



as compared to the function’s respective ln(x), for example,



and one can see an extremely interesting “coincidence”.  There is also a similar correspondence between the prime factorization of the denominators of the formulas for p = 2,3, and ln(x). 



III. Non-fundamental d with class number h(-d) = 1,2


Let k = ½, then,


p = 1:



p = 2:



p = 3



Evaluated as mln(x), these have a slightly different m from those with fundamental d.  If k = 0, then F(0)d = 1/π.



IV. Higher class numbers h(-d) = 4,6


A. Quadratic irrationals


Guillera pointed out that {A,B,C} may be algebraic and gave a few examples.  For fundamental d, we’ll use the smallest with class number h(-d) = 2, and the smallest d = {4u, 3v} with h(-d) = 4, namely d = {15, 68, 39}, respectively.  Let {φ, α, β} be the fundamental units,






All evaluate as mln(x) where x is an algebraic number of degree 2 but, as it turns out, the second one does not have the expected m = (1/2)4, but has m = (1/2)5, which will be relevant for the conjecture in the conclusion.



B. Cubic irrationals


We’ll use the smallest d with class number h(-d) = 3, and the smallest d = {4u, 3v} with h(-d) = 6, namely d = {23, 116, 87}. Let {x,y,z} be the real roots of,




respectively, then,



As expected, the general form is mln(x’) where x’ now is an algebraic number of degree 3.  And so on (apparently) for other d with higher class numbers.  As usual, all six formulas have F(0)d = 1/π.



V. Conclusion: Conjecture


Conjecture: “Given the function,


with the Pochhammer products hp as defined previously, {A,B,C} as algebraic numbers of degree N, and fundamental discriminant d.  If,






where {x, y} are algebraic numbers of degree N.” 


Some points:


1. Since ln(ab) = bln(a), the exponent 6 may be reduced to either {5, 4, 3} if x is a square, fourth power, or eighth power, respectively, over a field of the same degree N,  as was seen in the examples.


2. If indeed the “equalities” are true, there should be a closed-form expression for x as a function of d, presumably complicated, similar to how the Chudnovskys and Borweins derived {A,B,C} from d.


3. Furthermore, as Guillera pointed out, there might be interesting evaluations over other constant k, one of which he tried was k = ¼.  I tried negative values and found that, for k = -½, y apparently is an algebraic number that has the same degree as {A,B,C}.  Thus, we have,


F(-½)d = y = {-70.5, -248, -1752, -88056, -144088, -3107743896}


for d = {7, 11, 19, 43, 67, 163}, respectively.  I do not know what these integers are, as their factorizations are not divisible by d.  For d with higher class number, 


F(-½)d = y = {-15(89+45√5)/4,  -4(143+34√17),  -(31+13√13)/4}


for d = {3(5),  4(17),  3(13)} using the formulas given above.  If {A,B,C} are algebraic numbers of degree > 1, it seems the minimal polynomial of y and {A,B,C} share the same factor of the discriminant d.


For the background behind this article, see J. Guillera’s 2003 paper, “Series closely related to Ramanujan formulas for pi” in his homepage


P.S.  Some relaxing music from Enya...



-- END --   

© Oct 2010, Tito Piezas III

You can email author at tpiezas@gmail.com



Previous Page        Next Page